Hello all, I have some queries related to a drone (Multirotor) project that I'm currently working on. I think this is the right place to ask, as many of my connections are experts in the field of drones and UAVs. Here are a few questions that I need assistance with: • How can a drone identify checkpoints and take pictures of each checkpoint autonomously? • How can we identify and count objects of different shapes with drones? I just need an overview of the hardware, such as which flight controller I can prefer and which board I can use so that it can process and store captured image data.
Bharath ram S
7mo
Checkpoints can be marked with help of GPS and mission can be planned around those checkpoints or you can set up GCP's (ground control points) manually and make the drone take pics of these points autonomously for Aerial view, otherwise the drone can be set to take pics at regular intervals of 3-5 Secs, just like it does for mapping missions. Identify and counting - should make use of post processing software or build an AI system and train the drone to identify that object and take pictures of it. Using a pixhawk cube Flight control with mission planner and a good camera flying at a proper altitude of below 50-100ft depending on the mission will give you clear pictures.
Hi Vishal Bharti For your project, the Pixhawk 2.4.8 and the Pixhawk 2.1 are both great choices; pick one based on your budget. To do image processing on your drone, you'll need a companion computer, such as a Raspberry Pi or NVIDIA Jetson Nano. You may use MATLAB for object detection and counting (you can find instructions online). MATLAB's Computer Vision Toolbox offers an extensive collection of tools for using algorithms to recognize and classify objects of different shapes inside images captured by the drone's camera. Ensure that MATLAB and the Pixhawk aircraft controller can communicate with each other and exchange commands and telemetry data using MAVLink.
Great project you're embarking on! Drones can identify checkpoints and autonomously capture images by using GPS and computer vision systems programmed with machine learning algorithms. These systems can be trained to recognize specific waypoints and objects, allowing for precise navigation and task execution. For identifying and counting objects, drones typically employ a combination of sensors and cameras with advanced image recognition software. This software can differentiate and count objects based on shape, size, and color. Regarding hardware, flight controllers like the Pixhawk or DJI N3 are popular among professionals for their reliability and features. As for onboard processing, you might consider companion computers like the NVIDIA Jetson series for real-time image processing capabilities. For storage, ensure you have high-speed, high-capacity microSD cards or onboard SSDs, depending on your data requirements. Remember to ensure that your hardware choices are compatible with one another and can support the weight and power requirements of your drone setup. Best of luck with your project!
Rithin Raj
7mo
I guess I can help you a bit : 1.Try defining GPS coordinates for each checkpoint. These coordinates will serve as reference points for the drone's navigation system 2.Equip the drone with sensors such as GPS, IMU and possibly additional sensors like LiDAR or depth cameras for positioning and obstacle avoidance. 3.use computer vision algorithms to detect and recognize checkpoints using techniques like object detection and image classification, similar to point tracking. I hope this answers your first question !
Project Resource Analyst @ HQW Aerospace UK Ltd | Optimization and Innovation
7moHi, not an expert but have worked with image detection and small scale robots. The easiest method would be to incorporate GPS for the coordinates (that's basic ik). When in that location, then u can train your image detection model to detect a particular checkpoint. The easier it is to distinguish, the better will be it's confidence value. The same goes with counting objects, it depends on what object you want to track and is there any library available that captures that object or not. If yes then you are good to start working on the model else you would have to train the model by having loads of sample images. This would be the fundamentals of image processing that would be used in this. Best use case can be, cattle herd monitoring incorporating a drone etc. there are many use cases. But remember to care about the ethics behind it as well. There's loads of ethical and non-ethical uses of this. I think CV2 should help you set up the fundamentals quite well and then u can fine tune it. Best of luck.